wtf is ditto.site?

This repository is ditto.site, a tool that turns any public website URL into a working, self-contained TypeScript codebase. It works by capturing exactly what a browser renders on a given page, then generating a project from that capture, either a Next.js app by default or a Vite React app if you request it. The authors are careful to distinguish this from a normal git clone: you do not need the site's original source code or an existing repository at all, you just point the tool at a live URL and it writes you a fresh, runnable project based on what actually shows up in the browser. It is described as a capture-to-code pipeline rather than an AI page author, meaning the same frozen capture always produces the same generated app. There are three main ways to use it. First, a hosted REST API lets you submit a URL and get back either a finished job or a queued one you can poll, returning a file map of every generated file with its content and hash. A companion command line tool can then unpack that file map into an actual project folder on disk, or you can download the whole generated app as a single archive. Second, there is a hosted MCP server built specifically for AI agents, exposing tools to start a clone, check its status, list the generated files, and read only the specific files needed, rather than dumping everything into the agent's context at once. Third, you can run the whole thing locally by cloning this repository, installing dependencies, and running a clone command directly against any URL, which produces a ready to run app with clear pointers to the files that are safe to edit. The project also supports a full local service mode with Postgres and file storage through Docker Compose, for anyone who wants to queue and manage clone jobs themselves instead of using the hosted API. It is released under the MIT license and requires Node.js version 20 or newer.